From 173f889d2afc35a84de5babdae1d38fb2e7be884 Mon Sep 17 00:00:00 2001
From: "lancedb-gatefixer[bot]"
 <313497061+lancedb-gatefixer[bot]@users.noreply.github.com>
Date: Thu, 6 Aug 2026 16:44:20 +0800
Subject: [PATCH] test(python): cover stale scalar prefilters in hybrid search
 (#3865)

## Summary

- capture the stale-index state behind the reported fixed-size-binary
panic: the vector and FTS indices cover newer fragments while the BTree
prefilter does not
- verify vector, FTS, and hybrid searches return matches from both
scalar-indexed and unindexed fragments without panicking
- preserve binding-level coverage for the Lance fix in
https://github.com/lance-format/lance/pull/3768, which restricts
incomplete scalar prefilters when search indices are further ahead

The production root cause is in Lance and the current LanceDB dependency
already contains that fix, so this change adds the missing LanceDB
Python regression coverage.

## Validation

- `cd python && uv run --no-sync pytest
python/tests/test_hybrid_query.py::test_hybrid_query_with_stale_fixed_size_binary_prefilter
-q`
- `cd python && uv run --no-sync pytest
python/tests/test_hybrid_query.py -q`
- `python/.venv/bin/ruff check .`
- `python/.venv/bin/ruff format --check
python/python/tests/test_hybrid_query.py`

Fixes #2370

<!-- lance-gatekeeper-fix:v1 agent=5d16e59b9e513fd9247e0698732fa283
generation=1 -->

Co-authored-by: Gatefixer <313497061+lancedb-gatefixer[bot]@users.noreply.github.com>
---
 python/python/tests/test_hybrid_query.py | 82 +++++++++++++++++++++++-
 1 file changed, 81 insertions(+), 1 deletion(-)

diff --git a/python/python/tests/test_hybrid_query.py b/python/python/tests/test_hybrid_query.py
index 65a7890bf..72dcaaa49 100644
--- a/python/python/tests/test_hybrid_query.py
+++ b/python/python/tests/test_hybrid_query.py
@@ -12,7 +12,7 @@ import pyarrow.compute as pc
 import pytest
 import pytest_asyncio
 
-from lancedb.index import FTS
+from lancedb.index import BTree, FTS, IvfPq
 from lancedb.table import AsyncTable, Table
 
 
@@ -99,6 +99,86 @@ async def test_async_hybrid_query_filters(table: AsyncTable):
     assert result["text"].to_pylist() == ["cat", "b"]
 
 
+@pytest.mark.asyncio
+async def test_hybrid_query_with_stale_fixed_size_binary_prefilter(
+    tmpdir_factory,
+):
+    tmp_path = str(tmpdir_factory.mktemp("stale_scalar_prefilter"))
+    db = await lancedb.connect_async(tmp_path)
+
+    def fixed_size_binary(value: int) -> bytes:
+        return value.to_bytes(16, byteorder="big")
+
+    num_rows = 1000
+    data = pa.table(
+        {
+            "space_id": pa.array(
+                [fixed_size_binary(i) for i in range(num_rows)],
+                type=pa.binary(16),
+            ),
+            "text": ["book"] * num_rows,
+            "vector": pa.array(
+                [[float(i), float(i)] for i in range(num_rows)],
+                type=pa.list_(pa.float32(), 2),
+            ),
+        }
+    )
+    table = await db.create_table("test", data)
+    await table.create_index(
+        "vector", config=IvfPq(num_partitions=4, num_sub_vectors=2)
+    )
+    await table.create_index("space_id", config=BTree())
+    await table.create_index("text", config=FTS(with_position=False))
+
+    # Advance the search indices without advancing the scalar index. This is the
+    # state that previously let hybrid search use an incomplete scalar prefilter.
+    await table.add(data)
+    lance_dataset = await table.to_lance()
+    lance_dataset.optimize.optimize_indices(index_names=["vector_idx", "text_idx"])
+    await table.checkout_latest()
+
+    scalar_stats = await table.index_stats("space_id_idx")
+    assert scalar_stats is not None
+    assert scalar_stats.num_indexed_rows == num_rows
+    assert scalar_stats.num_unindexed_rows == num_rows
+
+    for index_name in ["vector_idx", "text_idx"]:
+        search_stats = await table.index_stats(index_name)
+        assert search_stats is not None
+        assert search_stats.num_indexed_rows == num_rows * 2
+        assert search_stats.num_unindexed_rows == 0
+
+    matching_ids = [5, 10, 15, 20, 25, 30]
+    literals = [
+        f"arrow_cast(0x{fixed_size_binary(i).hex()}, 'FixedSizeBinary(16)')"
+        for i in matching_ids
+    ]
+    predicate = f"space_id IN ({', '.join(literals)})"
+    expected_ids = sorted(fixed_size_binary(i) for i in matching_ids for _ in range(2))
+
+    vector_query = (
+        table.query().where(predicate).nearest_to([5.0, 5.0]).limit(num_rows * 2)
+    )
+    vector_results = await vector_query.to_arrow()
+    assert sorted(vector_results["space_id"].to_pylist()) == expected_ids
+
+    fts_query = (
+        table.query().where(predicate).nearest_to_text("book").limit(num_rows * 2)
+    )
+    fts_results = await fts_query.to_arrow()
+    assert sorted(fts_results["space_id"].to_pylist()) == expected_ids
+
+    hybrid_results = await (
+        table.query()
+        .where(predicate)
+        .nearest_to([5.0, 5.0])
+        .nearest_to_text("book")
+        .limit(num_rows * 2)
+        .to_arrow()
+    )
+    assert sorted(hybrid_results["space_id"].to_pylist()) == expected_ids
+
+
 @pytest.mark.asyncio
 async def test_async_hybrid_query_default_limit(table: AsyncTable):
     # add 10 new rows